Re: The wirehead problem

On Oct 7, 11:27 pm, c...@xxxxxxxx (Curt Welch) wrote:
Tim Tyler <seemy...@xxxxxxxxxxxxxx> wrote:
Curt Welch wrote:
How far down do you want to go?  Gas molecules?  Atoms?  Why stop there?

IMO, you don't need to go far down the rabbit hole, in the case of
a gas expanding to fill a box: the basic idea of maximising entropy
(as explained by Roderick Dewar, following Jaynes) works just fine.

My point is that you have to go down at least _one_ level in reduction to
create an answer to "why".

Absolutely not. This sort of idea is akin to strong eliminationism,
which isn't very productive. For example, do you really think that
all of the explanations and why questions and answers in biology are
wrong because they do not give them in terms of chemistry? So that,
for example, discussions of heredity of traits need to precisely
reference DNA to adequately describe, for example, why two people with
blue eyes will only produce children with blue eyes, but two people
with brown eyes can produce blue-eyed children?

Sounds like a nice why question to me, and dominance and recessiveness
seems to be an adequate answer, without ever leaving biology at all.

 You can't just give an effect a new name, and
say the new name is _why_ the effect exists (well, we can do it and we do
it all the time, but you haven't explained anything in the process other
than the link between the views).  There are many ways to use language to
create alternate views of the same effect without explaining why the effect
exists.

I'm actually quite surprised at your view here, because anyone who's
done programming at any level -- and fixing bugs -- will clearly
disagree that you need to reduce to answer why questions. For
example:

Let's say that I have a bug in the code. After investigating it, we
discover that the system is considering the object to be unchannelized
instead of channelized, and that's why the "bad behaviour" occurs.
Now, if someone asks me why that behaviour -- or bug -- appears, it's
a perfectly good answer to say "Because it thinks this thing is
unchannelized". I do not need to know the precise implementation of
the bug to give that answer. There could be multiple implementations
for why that is occurring -- wrong values in the database, a wrong or
missing condition in a function, an invalid interpretation, etc, etc.
I can fix the problem -- possibly -- without ever addressing the "root
cause" of the implementation; I can hack it, and solve the problem,
and still say that I know why the behaviour appeared. It's only when
I want to get to the lower cause that I have to go one step down, but
I don't need that to give a technically accurate answer to the why
question; I SHOULD, for various reasons, but I don't need to.

So, no, you can answer why questions without getting into how that why
is implemented.


We do it all the time in talking about human behavior for example.  We see
someone (ourselves) select cookies over dog *** as food, and we say "he
did it because he likes cookies better than dog ***".

Talking like that implies that the _cause_ of the behavior is our internal
"desire".  But that's just bull ***.  We use the word "like" to label the
behavior, not to label some real internal cause of the behavior.  "desire"
is not "why" we eat cookies.  "desire" is the odd backassward way we label
behavior which shows a statical bias in selecting cookies over dog ***.

Except that human behaviour contradicts this naive notion. For
example, someone may indeed select dog *** over cookies because they
feel that they are worthless and don't deserve cookies. If you
insisted that that meant that they "liked" cookies, they'd get very
upset with you and point out that, no, they DON'T like cookies. So
your comments about how we use the word does not align with how we use
the word.

So, in general, saying that someone takes cookies because they like
them IS the answer to the why question, and to the causal question,
and the one that we are looking for at that level. We don't need to
know how "likes" are implemented (and/or learned or formed) to do a
perfectly valid causal analysis or to answer the why question; they
could be behaviourist, functional, neurons, computer chips or whatever
and the answer is still the same.

You want to drag us into the implementation, and I ain't buyin' it. I
could write a system with folk psychological wants and it would
fulfill the requirements of an AI system just as well -- and probably
better -- than your behaviourist model.


To say that entropy is the cause of gas filling the volume is no different
than trying to say entropy is the cause of entropy, or gas filling the
volume is the cause of gas filling a volume, or gas filling a volume is the
cause of entropy.

Saying these things helps us understand the relationship between the
concepts, but doesn't in fact _explain_ anything about the universe other
than the relationship between the language concepts.

Well, here again, you have a problem. Let me put the relevant
question this way: I have a pot of boiling water with a lid on it.
The lid starts to jump and the water splashes over the side, and I ask
"Why does that happen?". And someone says that since the air is
getting heated, it expands, and that pushes the lid off to relieve
some of the pressure, and also pushes the water out over the side.
What lower level am I appealing to here? Certainly not molecules, or
molecular movement, or even entropy. Is this not a perfectly
acceptable answer to the "Why" question I asked? Seems to me that it
is. Only if I want to know WHY gas expands to I have to start
appealing to lower levels. But that's a implementation question; it
wouldn't matter to my first question if the cause was molecular
movement or little gremlins pushing the air around.

To me, it seems to be saying something the kin of, "The acceleration due to
gravity is _why_ the apple increases in speed as it falls".

That is a bad example, because the answer is contained in the
statement: the apple increases in speed because it accelerates
basically translates DIRECTLY to "It accelerates because it
accelerates".

Answering "An apple increases in speed as it falls because of gravity"
is actually the answer here.


The concept of acceleration does not explain why the apple falls or why
there is gravity or why gravity works the way it does.  It was the fall of
the apple that was used to define what gravity is.  Gravity is just a
different name for the way in which things fall to the ground.  Gravity is
not _why_ they fall (even though as I point out, we like to talk as if it
is).

Well, actually, gravity IS why they fall ... but gravity is an
attraction between two massed objects as part of its definition, which
is why that IS the explanation.


As far as I know, gravity is a fundamental force of the universe which has
never been reduced to something simpler, as such, we don't have a "why"
answer for it.  We don't know why apples fall like they do, we just know
they do and we call it a fundamental force of the universe, we call the way
they fall "gravity".

Um, this is a complete misinterpretation of gravity, as you'd see if
you did Astrophysics where gravity is, in fact, VERY real and
unrelated to actual falling in any way.

The issue with Dark Matter, for example, is the result of
gravitational effects and the discrepancies in them.


.

Loading